Technical Field
[0001]
The present invention relates to an image processing
device that recognizes a surrounding environment of a
vehicle by utilizing a sensor mounted on the vehicle.
Background Art
[0002]
PTL 1 describes a technique for detecting a road
shoulder by using a sensor and using the road shoulder for
preventive safety and automatic driving.
Citation List
Patent Literature
[0003]
PTL 1: JP 2009-053818 A
Summary of Invention
Technical Problem
[0004]
In recent years, warnings and control functions
related to preventive safety have reached a widespread
stage. Development of a sensing function for predicting an
3
accident in advance is also accelerated in order to realize
the preventive safety. Currently, many of the widely used
functions relate to preventive safety functions on an
expressway, driving assistance, or an emergency brake for
preventing a collision in advance on a general road. Among
these functions, a traveling road deviation prevention
function by road end detection which is one of the
preventive safety functions has not yet reached the
widespread stage since a degree of difficulty of sensing
the road end is high.
[0005]
Unlike a lane that is artificially painted to
indicate a section of a traveling road, a road end does not
necessarily have a wall of an artifact or a step of a
sidewalk. For example, in the case of the expressway, an
artificially maintained wall is mostly provided in the road
end, and the road end is a continuous and clean curve even
though the wall is not as good as the lane in many cases,
and recognition of the road end is relatively easy. On the
other hand, in the case of the general road, an obstacle
such as a utility pole, a rock, or grass may run out onto
the traveling road, and it is difficult to model a
curvature shape that facilitates traveling of a vehicle in
advance, and a degree of difficulty of sensing to detect
the road end is high. In the general road, not only a
4
shape of the road end but also discontinuity of the road
end and a classification of the road end are mixed in many
cases, and there are many scenes where it is difficult to
recognize the road end.
[0006]
In the general road, structures (objects) such as
fences, guardrails, and sidewalks are connected
continuously even though a shape in a direction along the
traveling road is basically not a somewhat smooth curve in
many cases. However, even in these structures,
discontinuity occurs at an entrance to a store or home, or
other obstacles block the traveling road, and the
classification of the road end may be immediately changed.
[0007]
On a side of the traveling road, there are a region
of lawn, gravel, or soil without pavement which is very
short and has no difference in height from the road surface,
a region having a height lower than the road surface, such
as a side groove, a rice field, or an upland field, and
boundaries between these regions and the traveling road are
also the road ends. As described above, not only a
classification of the region varies, but also various
objects such as a utility pole, an obstacle, grass, or a
lid of a side groove, and a tree is discontinuously present
at the road end of the traveling road, and it is an
5
important problem to more appropriately recognize these
objects.
[0008]
The present invention has been made in view of the
above problems, and an object of the present invention is
to provide an image processing device that senses a
surrounding environment of a host vehicle by using a sensor
and outputs information necessary for control and a warning
of the host vehicle in order to control the vehicle more
safely and comfortably.

WE CLAIM:
[Claim 1]
An image processing device that detects a road end of
a traveling road based on a captured image of an in-vehicle
camera, wherein the image processing device
recognizes classifications of objects constituting
the road end from the captured image, and
changes a parameter for detecting the road end in
accordance with the classification of the object.
[Claim 2]
The image processing device according to claim 1,
wherein
the in-vehicle camera is a stereo camera, and
the image processing device
votes features of the objects obtained from the
stereo camera on a map in accordance with a behavior of a
host vehicle, and
performs detection of the road end by using a
voting result on the map.
[Claim 3]
The image processing device according to claim 1,
wherein the image processing device
recognizes a plurality of classifications of the
objects, and
performs detection of the road end by simultaneously
66
using the objects of the plurality of classifications.
[Claim 4]
The image processing device according to claim 3,
wherein the image processing device
extracts the objects of the same classification from
among a plurality of the objects recognized from the
captured image, and
performs the detection of the road end by connecting
features of the objects of the same classification.
[Claim 5]
The image processing device according to claim 1,
wherein the image processing device recognizes, as one
classification of the object, a step between a region
present at a position lower than a road surface of the
traveling road and the traveling road from the captured
image.
[Claim 6]
The image processing device according to claim 1,
wherein the image processing device recognizes, as one
classification of the object, a boundary between the
traveling road and a region present at a position having
the same height position as a road surface of the traveling
road by using texture information of the captured image.